# Anras573 Aichemist Tool Preferences > Tool Selection Preferences - Skill: `tomevault-io/anras573-aichemist-tool-preferences` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/anras573-aichemist-tool-preferences` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/anras573-aichemist-tool-preferences/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/anras573-aichemist-tool-preferences --- # Tool Selection Preferences This skill provides guidance on choosing between equivalent tools when multiple options are available. Following these preferences ensures consistent behavior, better performance, and predictable output across all agents. ## GitHub Operations **Prefer `gh` CLI (via Bash) over GitHub MCP tools.** | Operation | Preferred | Avoid | |-----------|-----------|-------| | View PR details | `gh pr view` | GitHub MCP `pull_request_read` | | Get PR diff | `gh pr diff` | GitHub MCP `get_commit` | | List PRs | `gh pr list` | GitHub MCP `list_pull_requests` | | Search PRs | `gh pr list --search "query"` | GitHub MCP `search_pull_requests` | | Post PR comment | `gh pr comment` | GitHub MCP `add_issue_comment` | | View issues | `gh issue view` | GitHub MCP `issue_read` | | List issues | `gh issue list` | GitHub MCP `list_issues` | | Create issues | `gh issue create` | GitHub MCP `issue_write` | | Search code | `gh search code "query"` | GitHub MCP `search_code` | ### Why Prefer `gh` CLI 1. **Already authenticated** - Uses user's existing git/GitHub credentials 2. **Less indirection** - Avoids an extra MCP/tooling hop by talking directly to GitHub APIs 3. **Predictable output** - Consistent format with `--json` flag 4. **Better errors** - Clear, actionable error messages 5. **Uses local repo state** - Can operate directly on the checked-out repository when appropriate ### When GitHub MCP Tools ARE Appropriate Use GitHub MCP tools only when they provide functionality the CLI lacks: | Use Case | Reason | |----------|--------| | Inline PR review comments | CLI doesn't support line-specific review comments | | Pending review management | Creating/submitting pending reviews with multiple comments | | File contents at specific ref | When you need content without cloning the repo | ### Common `gh` CLI Patterns ```bash # View PR with specific fields gh pr view --json number,title,body,baseRefName,headRefName # Get PR diff gh pr diff # List PRs with filters gh pr list --state open --author @me # View issue details gh issue view 123 --json title,body,state,labels # Search issues gh issue list --search "bug in:title" # Post comment to PR gh pr comment 123 --body "Comment text" # Get repo info gh repo view --json nameWithOwner,defaultBranchRef # Search code across repos gh search code "pattern" --repo owner/repo ``` ## Git Operations **Prefer native git commands over any wrapper or MCP tool.** | Operation | Preferred | Avoid | |-----------|-----------|-------| | Check status | `git status` | Any wrapper | | View diff | `git diff` | GitHub MCP diff tools | | Create branch | `git checkout -b name` | GitHub MCP `create_branch` | | Commit changes | `git commit` | Any remote commit API | | Push changes | `git push` | GitHub MCP `push_files` | ### Rationale Git operations should be local-first. Using MCP tools for git operations: - Bypasses local hooks (pre-commit, pre-push) - Doesn't update local repository state - May cause sync issues between local and remote ## Atlassian/Jira Operations **Prefer Atlassian MCP tools (the supported integration in this repo).** The Atlassian MCP server (`atlassian/*` tools) is the preferred choice for Jira and Confluence operations in this environment. See the Jira skill for detailed guidance. ## Browser Automation **Prefer `playwright-cli` (via Bash) over Playwright MCP.** | Reason | Detail | |--------|--------| | Token-efficient | Avoids loading large tool schemas and accessibility trees into context | | Better for coding agents | Concise CLI commands fit within limited context windows alongside codebases | ### When Playwright MCP IS Appropriate - Exploratory automation requiring persistent state and rich introspection - Long-running autonomous workflows where continuous browser context outweighs token cost ### Common playwright-cli Patterns ```bash # Open a page and snapshot for element refs playwright-cli open https://example.com playwright-cli snapshot # Interact using refs from the snapshot playwright-cli fill e12 "text" playwright-cli click e5 playwright-cli press Enter # Capture output playwright-cli screenshot --filename=result.png playwright-cli console error # Named sessions for parallel work playwright-cli -s=myapp open https://example.com ``` ## Documentation Lookups **Prefer MCP tools** for documentation: | Source | Tool | |--------|------| | Library docs | Context7 (`context7/*`) | | .NET/Microsoft docs | Microsoft Learn (`microsoft-docs/*`) | These MCP tools provide curated, up-to-date documentation that's more reliable than web searches. ## Decision Framework When unsure which tool to use, apply this framework: ``` 1. Is this a documentation lookup? YES → Prefer MCP documentation tools (Context7, Microsoft Learn) NO → Continue to step 2 2. Is there a local CLI tool that does this? YES → Use the CLI (gh, git, npm, dotnet, etc.) NO → Continue to step 3 3. Does the MCP tool provide unique functionality? YES → Use the MCP tool NO → Prefer CLI if available 4. Does the operation need to respect local state/hooks? YES → Must use local tools (git commit, npm install) NO → Either is acceptable ``` --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/anras573) — claim your Tome and manage your conversions.